Microsoft Build 2025| The main Highlights

Microsoft is once again bringing together the community of developers, software engineers, IT professionals, and enterprises to launch the latest news of its services and platforms. Microsoft Build is back, and we have all the details.

*We will update this post with daily releases and news.

This is the second year since the pandemic that the event has regained its face-to-face essence and offers the opportunity to attend in person at the Seattle Convention Center.

However, given the great reception of recent years thanks to greater accessibility, the event will also be streamed online and free to anyone who registers. The dates chosen are May 19, 20, and 21.

This edition will be full of new innovative proposals and sessions ranging from the latest innovations in code development or challenges, to case studies from world experts.

The sessions will be led by Satya Nadella (CEO of Microsoft) and leading industry figures and partners such as Jay Parikh, Charles Lamanna, Scott Guthrie, Asha Sharma, Kevin Scott, and Amanda Silver, among many others. In addition, they will be organized in several virtual modalities to connect attendees with Microsoft engineers, experts, and other presenters during the meeting.

Recap Day 1

As usual, the kickoff of the event has been a big boom of launches, news, and surprising announcements, especially focused on AI agents and the creation of an open agentic network.

Satya Nadella, CEO of Microsoft, was again in charge of taking the reins on this first day, and in this video, you can see his Keynote, with the most outstanding moments:

Microsoft 365 Copilot Tuning

With this service, customers can use their own business knowledge to train models that perform domain-specific tasks with greater speed and accuracy.

Copilot Tuning ensures that access to an optimized model is restricted to the permissions of the underlying training data within the enterprise.

Agents created with Microsoft 365 Copilot’s Agent Builder can leverage these optimized models for a wide variety of everyday enterprise tasks.

Microsoft Copilot Studio

This service expands with new features that make it even easier for creators to leverage professional coding experiences to create. These are characterized by greater flexibility and advanced tools to create more powerful and secure agents, scalable across multiple channels:

• The Agents for Visual Studio toolkit optimizes enterprise agent development by integrating AI tools such as Agents SDK and Azure AI Foundry.
• The SDK enables the creation of scalable, enterprise-grade, multi-channel agents, as well as offering flexibility for advanced needs and easy integration with Copilot Studio and Visual Studio, as well as the ability to publish to multiple platforms with more than 10 messaging channels.
• Copilot 365 APIs enable developers to create fast, contextual, secure, and data-driven generative AI experiences from Microsoft 365, with recovery, chat, and compliance functionality.
• Support for the Model Context Protocol in Copilot Studio ensures that agents have consistent, controlled access to data and models across external systems.

GitHub Copilot and GitHub Models coding agent

GitHub Copilot evolves from an assistant to an agentic AI partner with an asynchronous coding agent, the first of its kind, integrated into the platform.

Request management, lightweight evaluations, and enterprise controls have also been added to GitHub Models, with the goal of allowing teams to experiment with best-of-breed models without leaving GitHub.

And, another new feature is that it also opens up GitHub’s open source Copilot Chat VS Code, where AI-driven capabilities will now be part of the same open source repository that powers the tool.

Windows AI Foundry

Introduced as an evolution of Windows Copilot Runtime, it offers a unified and reliable platform that supports the AI developer lifecycle, from model selection, optimization, tuning, and deployment to client and cloud.

AI Foundry now includes several features:

• Windows ML is the foundation of the AI platform and the integrated AI inference runtime environment, allowing developers to incorporate their own models and efficiently deploy them across the entire partner ecosystem.
• Integrates Foundry Local and other model catalogs such as NVIDIA’s Ollama and NIM, giving developers quick access to open-source models ready for use on a variety of Windows systems.
• It offers out-of-the-box AI APIs based on Windows inbox models in PC Copilot+ for key language and vision tasks. Also, new capabilities such as LoRA for optimizing the inbox SLM, Phi Silica, with custom data. As well as new APIs for semantic search and knowledge retrieval.
• Integrates enhanced security capabilities, helping organizations to securely deploy generative AI applications.

New features of Copilot Pages

They have introduced a completely new workflow, designed for the Copilot era, that allows users to turn a Copilot response into a dynamic, editable and shareable page. New features include:

• Create pages from a cell phone.
• Convert a page into a Word document with a single click.
• Find pages more easily with the ability to search and filter a list of pages.
• More interactivity and output types are available with interactive graphics and code blocks.

Azure Migrate

Semi-structured data is increasingly crucial in the era of AI applications, providing the structure needed for AI algorithms to learn from data and understand complex information.

The unification of data estates expands with the preview release of Cosmos DB (NoSQL) on Fabric, which will offer enterprise-grade dynamic scalability, consistent reliability, and low latency for semi-structured data provisioning.

Digital Twin Generator in Microsoft Fabric

Having a digital twin can significantly improve an organization’s ability to make informed decisions by integrating the physical world with the digital world.

Fabric’s digital twin generator is designed to simplify and streamline the process, enabling organizations to leverage the potential of their data.

It will be a new Real-Time Intelligence feature in Fabric, already in preview, that will provide a quick and easy way to create and manage digital representations of real-world environments.

Grok AI comes to Azure

Nadella announced that Grok (Grok 3 and Grok 3 Mini) and Mistral are integrated into AI Foundry. With this, Microsoft enables all major language models to be used in its ecosystem: OpenAI’s GPT, Meta’s Llama, DeepSeek, and now Mistral and Grok.

Developed by xAI , it is a family of large language models designed to push the boundaries of AI innovation and accelerate human scientific discovery. According to xAI , Grok 3, the flagship model, combines exceptional reasoning with vast pre-training knowledge, honed on xAI’s Colossus supercluster, with ten times the computational power of previous leading models. Grok 3 models excel in reasoning, mathematics, coding, world knowledge and instruction following, setting a new standard for enterprise AI.

Grok 3 now appears in the catalog alongside other core models. As with an OpenAI or Hugging Face model, they can quickly find the Grok 3 model card (with details and benchmarks) and implement it with just a few clicks.

Introduction to DevOps with Agents

Microsoft is evolving the software development lifecycle through autonomous and semi-autonomous agents that operate as members of development and operations teams, automating, optimizing, and accelerating each stage of the software lifecycle.

The new agent capabilities in GitHub Copilot aim to solve routine and complex tasks, accelerate application time to market, increase code quality and security, reduce technical debt, and redefine the economics of operating, maintaining, and modernizing applications in production.

With agent DevOps, developers will be free to focus on higher-value creative work, while operators will be able to proactively identify, mitigate,e and resolve problems in production.

New capabilities include: GitHub Copilot scheduling agent, new application modernization capabilities, new SER agent, GitHub models, agent mode, etc.

SDK of Microsoft Purview

For enterprises to confidently adopt AI applications, developers must build their AI solutions with enterprise-grade data security and compliance controls.

The Microsoft Purview SDK, in preview, will provide REST APIs, documentation, and code samples for integrating Microsoft Purview data security and compliance into AI applications directly from any integrated development environment.

By integrating Microsoft Purview REST APIs into AI applications, code will send data related to requests and responses to Microsoft Purview, which automatically sends signals to security and compliance teams to support investigations through discovery, protection, and data management.

In addition, Microsoft Purview performs real-time classification of requests to identify and block access to sensitive data by an extensive language model. This can empower organizations and data security administrators by facilitating the adoption of AI with industry-leading data security and compliance capabilities.

Agentic Web Support

Microsoft and GitHub have joined the Model Context Protocol (MCP) Steering Committee and are providing extensive proprietary support for MCP across their entire platform of agents and frameworks, spanning GitHub, Copilot Studio, Dynamics 365, Azure, Azure AI Foundry, Semantic Kernel, Foundry Agents, and Windows 11.

Microsoft and GitHub announce two new contributions to the MCP ecosystem to help promote secure, large-scale adoption of the open protocol:

• Identity and Authorization Specification: Microsoft’s identity and security teams have collaborated with Anthropic, the MCP Steering Committee, and the MCP community to design an updated authorization specification that enables MCP-connected applications to more securely connect to servers. This specification allows users to use their Microsoft Login ID or other trusted login methods for LLM-based agents and applications to access data and services such as personal storage drives or subscription services.
• Public, community-managed MCP server registry: This service allows anyone to deploy public or private, up-to-date, centralized repositories for MCP server entries, and enables the discovery and management of various MCP deployments with their associated metadata, configurations, and capabilities.

Microsoft Planetary Computer Pro

In the preview release, it will enable customers to deploy it in Azure and help them streamline the generation of geospatial information and integrate it into enterprise data and AI workflows.

It will facilitate the ingestion, management, and dissemination of customers’ private geospatial datasets, as well as facilitate the integration of this data with business operations for more informed decision making.

Microsoft Discovery

It is an enterprise AI platform that helps accelerate research and discovery by transforming the entire process with artificial intelligence, from scientific reasoning to hypothesis formulation, candidate generation, simulation, and analysis.

It is a platform that enables scientists and researchers to collaborate with a team of specialized AI agents to drive scientific results with speed, scalability, and accuracy, using the latest innovations in AI and supercomputing.

Power Apps

It introduces powerful new features for building AI-based solutions that enable closer collaboration with agents for both builders and users:

• Solution Workspace: developers collaborate with Agentic AI to define business requirements, generate data models, process maps and design the solution architecture.
• The agent feed for apps is a new hub for agent and human collaboration, where users can view, manage, and monitor agents within their app.
• Incorporating code-first applications into Power Platform will allow developers to develop and deploy applications with their favorite tools, with full control of their code, and then import them into Power Platform.

w features announced at Build and much more! Get your tickets now!

Microsoft Edge

With the new AI APIs in Microsoft Edge, developers can now incorporate AI capabilities into their web applications quickly and easily using in-browser models.

In addition, these APIs allow websites and extensions to access Phi-4-mini, a 3.8 billion parameter model developed by Microsoft that compares favorably to significantly larger models in testing.

This new family includes the Prompt API for easily triggering the model and a set of writing assistance APIs for generating, summarizing and editing text. As well as the Translator API for text translations, available soon.

Windows 11

Evolves into the future of agency with native Model Context Protocol (MCP) support, providing a standardized framework for AI agents to connect with native Windows applications, enabling them to seamlessly participate in agency interactions.

Available soon in preview, it will provide the ability for Windows applications to offer specific functionality to extend the skills and capabilities of agents installed locally on a Windows PC.

Azure Databricks

Developers will be able to use Azure AI Foundry to easily create AI agents that leverage their valuable enterprise data stored in Microsoft Azure Databricks.

This new feature, in preview, will enable AI agents to connect to Azure Databricks for real-time data processing, making it easier to design and customize reliable, enterprise-ready AI agents.

Going a step further, the connector will enable agents to leverage Spark jobs from data scientists in Azure Databricks. In addition, they will be able to interact with data through natural language queries (NLQs) via Genie rooms in Azure Databricks, facilitating a multi-agent workflow.

Chat in Power BI and Copilot Studio

Chat with data experts will enable teams to explore and reason about complex data sets by simply asking questions. This will remove barriers between users and information, enabling everyone from business analysts to data scientists to make faster, more confident decisions directly from the tools they already use.

Now included are two new experiences that will enable Power BI and Copilot Studio users to interact with their data, expanding the ways they can explore, analyze, and act on it, including:

• Chat with your data: this Copilot experience will be easily accessible from the left pane to help you find relevant reports, analyze data, and answer questions for any data that users have access to across multiple reports, semantic models, applications, and data agents, without having to navigate to specific reports.
• Enrich Copilot Studio custom agents with information from Fabric: these agents will be deployed in channels such as Microsoft Teams and Microsoft 365 Copilot. Once connected, the custom agent will use Fabric’s data agent to pull information from OneLake, while respecting data access permissions. Developers will also be able to define actions to automate processes, making it easier for users to interact with data and streamline tasks without leaving the chat.

SQL Server 2025

In the preliminary version, it will accelerate time to market for new customer applications because they will be able to:

• Increase search intelligence using advanced semantic search along with full-text search and filtering.
• Process and manage data streams more simply and efficiently using native JSON support, built-in REST API, Change Event Streaming for real-time data updates, and GitHub Copilot.
• Leverage the more secure database to improve credential management and reduce potential vulnerabilities with support for Microsoft Entra ID managed identities through Azure Arc.
• Increase workload uptime and improve concurrency of SQL Server applications with enhanced query optimization, optimized locking, and improved failover reliability.

Microsoft Defender

Microsoft Defender for Cloud, Microsoft’s cloud-native application protection platform (CNAPP), will integrate AI security posture and runtime threat protection for AI services into the Azure AI Foundry portal.

With this integration, developers will gain visibility into security risks, vulnerabilities, and security alerts from Defender for Cloud to protect against threats, such as jailbreak attacks, sensitive data leaks, and wallet abuse, within the Azure AI Foundry environment.

Microsoft Enters Agent ID

Microsoft reaches the first crucial milestone in extending its complete AI agent identity and access solution with the introduction of Microsoft Entra Agent ID. Currently in preview, this solution will focus on proprietary integration with Microsoft Copilot Studio and Azure AI Foundry.

It will address the problem of AI agent proliferation by assigning a unique identifier to each agent in an environment. With Entra Agent ID, administrators will be able to:

• See all AI agents created with Copilot Studio and Azure AI Foundry in one place.
• Know what those agents can access within the organization.

Stay tuned as we continue to recap the highlights of Microsoft Build 2025.

Don’t miss dotNET 2025, where Microsoft leaders like Scott Hunter, Glenn Condron, or James Montemagno will talk about the new features announced at Build and much more! Get your tickets now!